• DocumentCode
    379076
  • Title

    Improving SMT performance scheduling processes

  • Author

    Goncalves, Ricardo ; Navaux, Philippe

  • Author_Institution
    Departamento de Informatica, Univ. Estadual de Maringa, Brazil
  • fYear
    2002
  • fDate
    2002
  • Firstpage
    327
  • Lastpage
    334
  • Abstract
    Nowadays, SMT (simultaneous multithreaded) architectures use aggressive techniques to execute instructions from different threads on shared resources simultaneously. SEMPRE (Superscalar architEcture with Multiple PRocesses in Execution) is a type of SMT architecture which was proposed to schedule and execute processes simultaneously. The waste of time on both process scheduling and context switching is minimal, providing high performance during the execution of applications. The SEMPRE architecture was analyzed and evaluated using execution-driven simulations of the SPEC benchmark suite. The simulations showed that process scheduling by hardware can provide reasonable performance over process scheduling by the operating system on equivalent SMT architectures, with little extra hardware. This higher performance is achieved because the hardware makes better use of the process time-slice. The performance of SEMPRE is always higher than the performance of traditional SMT, achieving more than 21% in some cases
  • Keywords
    multi-threading; parallel architectures; performance evaluation; processor scheduling; SEMPRE; SMT performance scheduling processes; SPEC benchmark suite; aggressive instruction execution techniques; context switching; execution-driven simulation; multiple process execution; operating system; process scheduling; process time-slice; shared resources; simultaneous multithreaded architectures; superscalar architecture; Computer architecture; Hardware; Operating systems; Parallel processing; Pipelines; Process control; Read only memory; Surface-mount technology; Switches; Yarn;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Parallel, Distributed and Network-based Processing, 2002. Proceedings. 10th Euromicro Workshop on
  • Conference_Location
    Canary Islands
  • Print_ISBN
    0-7695-1444-8
  • Type

    conf

  • DOI
    10.1109/EMPDP.2002.994302
  • Filename
    994302